Sexism in science is no secret, and the #MeToo movement is exposing the toxic environment that women in STEM experience.
A 2018 report by the National Academies of Sciences, Engineering, and Medicine revealed that sexism pervades all of academic science. In 2015, the Harvard Business Review published five biases pushing women out of STEM — including having to prove themselves repeatedly; walking a “tightrope” between masculinity and femininity; being perceived as incompetent after having children; a tug-of-war with other women; and isolation. Women deal with these biases in a number of ways, including using “whisper networks'' as a way to avoid toxic principal investigators (PIs) or laboratory directors. For example, graduate students may warn each other about a particularly sexist advisor or one who assigns too much work while taking the credit.
Whisper networks were created to help students navigate the system. Still, we should be working to end sexism in the sciences rather than asking women to try to work around it. Strangely, a new study published by Nature Communications reinforces sexism as the norm. They conducted a correlational analysis of approximately three million mentor–protégé pairs in various scientific fields to determine the role that the gender of one’s advisor plays in their later publication impact. The study uses high-tech algorithms to try to make sense of data scraped from published academic papers on the internet in a variety of fields. Rather than looking at actual mentor–protégé pairs, the study analyzes relationships between co-authors without having insight into whether two coauthors are in a mentor–protégé relationship.
The article concludes that women mentees (such as undergraduates or graduate students, but also postdocs and junior faculty) who wish to be successful in science should consider working with PIs that are men rather than “equally impactful” PIs that are women. Furthermore, the paper argues that because women PIs are cited less than men PIs, women PIs should avoid working together as this magnifies gender differences between men and women PIs.
In academic science, the impact factor of a journal determines its influence. Nature Communications had an impact factor of 12.21 in 2019, which means that papers published that year were cited about twelve times that year. Nature papers are often read and relied upon as the foundation to advance science. Nature Communications is part of the Nature publication family — Nature itself is one of the biggest-name journals in science, with an impact factor of about 43 in 2019. Getting a paper published in Nature is a huge achievement — so the fact that Nature Communications has endorsed working with men PIs over women PIs seems like a mandate from science itself.
The Nature Communications paper concludes with policy recommendations to rethink gender equity programs in academia using the insights published in their paper, without recognizing the sexist citation gap already inherent in academia.
Nature announced on November 19, 2020, that they are “investigating the concerns raised” regarding the paper, especially regarding the fact that “gender plays a role in the success of mentoring relationships between junior and senior researchers, in a way that undermines the role of women mentors and mentees.” On December 21, 2020, Nature retracted the controversial paper. In a retraction note posted on the Nature Communications website, the authors state, “We are an interdisciplinary team of scientists with an unwavering commitment to gender equity, and a dedication to scientific integrity. Our work was designed to understand factors that influence the scientific impact of those who advance in research careers. We feel deep regret that the publication of our research has both caused pain on an individual level and triggered such a profound response among many in the scientific community.”
Indeed, one of the main factors that contributed to the retraction of this paper was widespread criticism from the scientific community. Many scientists felt that the paper did not live up to the rigor expected of a Nature paper. Immediately after its publication, the paper galvanized scientists to write their own critique of the paper. A preprint rebuttal submitted to Nature states regarding the findings, “None of this is surprising; it is consistent with a wide body of research documenting pernicious systematic and structural gender biases.”
Mona Xu, an Associate Professor at Idaho State University, is the corresponding author of the rebuttal paper. “Many of us have been astonished that this got through peer review,” she told me. “To their credit, [Nature Communications does] something that many journals don’t do — they publish the review process.”
Checking out the peer review file and the discussion between the study’s authors and the reviewers, it is clear that there was a lot of criticism that the study’s authors were asked to address. Even after that, two out of four reviewers recommended against publication in the paper’s final state. Of the two who recommended publication, one wrote, “I think the choices made are reasonable even if not always correct (if there is such a thing in these cases).” So, only one out of four reviewers unequivocally supported the paper’s publication after several rounds of revisions.
Three out of four reviewers opted to stay anonymous, while one reviewer, Daniel Acuña, chose to reveal his identity. After the article’s publication, Acuña took to Twitter to discuss his disappointment. “I was surprised the editor continued to pursue this article after what I thought were major criticisms from many of us…” he tweeted.
Dr. Xu believes that such papers are damaging to the public’s perception of women in science just as much as they are harmful to science itself. “People may hear about the paper or read just the abstract without understanding how deeply flawed the study is, and erroneously believe it provides evidence that women are worse scientists and mentors. And for women in science, there are already so many barriers. Now there’s this message, being communicated on Nature’s huge platform, that if you’re a woman you shouldn’t work with other women. And even a suggestion that policies should be changed to dissuade women from mentoring and collaborating.”
Dr. Xu and colleagues’ criticisms of the paper include the problematic definition of mentorship as co-authorship; the simplistic definition of success as the number of one’s publications and citations; and the use of algorithms to determine gender, an error-prone process that led the authors to exclude over half their dataset. “It’s problematic to use an algorithm to presume gender and force a binary categorization, from names. But even if we ignore that, it’s irresponsible to present this supposed gender effect and claim causation when the results can be easily explained by systemic issues in academia and science,” Dr. Xu said.
It’s one thing to call out sexism in science, but it’s entirely another to make policy recommendations based on correlational relationships interpreted as causal ones. Sadly, though, this isn’t the first time Nature Communications has passed off correlation as causation. In September 2020, they published an article called “Tracking historical changes in trustworthiness using machine learning analyses of facial cues in paintings” that claimed to rediscover racism and eugenics. Just a few days after publication, the editors issued a notice that the paper was under review, but the journal has not retracted this article.
Those who are practitioners of science should be able to easily differentiate correlation from causation. While the paper’s retraction is a victory for both women in science and peer review, the original publication of this paper should be highly problematic for the scientific community. If anything, the fact that any analysis of the data could show that one could be more successful abandoning women PIs and graduate students — should serve as an eye-opener for everyone in academia, regardless of their gender.